Prospective Assessment of Alignment Between Multimodal Artificial Intelligence and Multidisciplinary Team Decisions in Gastrointestinal Oncology

Inclusion criteria

• Aged 18 years or older, with no restriction based on sex.
• Histologically confirmed gastric, colon, or rectal cancer; or highly suspected gastric, colon, or rectal malignancy based on clinical, endoscopic, or imaging findings, with further diagnostic, staging, or treatment planning required.
• Scheduled for multidisciplinary team evaluation at the study center because of an initial diagnosis, clinical staging, perioperative treatment, surgical planning, recurrence or metastasis, conversion therapy, treatment response assessment, or another complex clinical management question.
• A clearly identifiable clinical decision time point can be established, and the clinical information available before that time point can be defined.
• At the clinical decision time point, at least the basic medical history, the clinical question to be addressed, and at least one core source of information from endoscopy, pathology, or imaging are available. Completion of all examinations is not required, because missing examinations may be evaluated as part of diagnostic task planning.
• A corresponding multidisciplinary team recommendation regarding diagnostic tasks or the principal treatment decision can be obtained for comparison with the artificial intelligence output.
• Willing and able to understand the study and provide dated written informed consent.

Exclusion criteria

• Further examination confirms a condition other than gastric, colon, or rectal cancer, and the case is not applicable to the gastrointestinal oncology pathways evaluated in this study.
• The participant is undergoing routine follow-up or continuation of a previously established treatment plan and has no clinical question requiring additional diagnostic tasks, modification of the principal treatment pathway, or multidisciplinary team decision-making.
• A clear clinical decision time point cannot be established, or information available before and after the decision cannot be distinguished, preventing comparison of the artificial intelligence system and the multidisciplinary team under the same information conditions.
• Core clinical information is severely incomplete, such that the current stage of care and the clinical question cannot be identified and a clinically meaningful artificial intelligence output cannot be generated.
• A clear multidisciplinary team recommendation corresponding to the clinical decision time point cannot be obtained, or the multidisciplinary team documentation is insufficient for paired evaluation.
• The participant requires immediate resuscitation or urgent clinical management, and study recruitment could interfere with necessary medical care.
• The participant is unable to provide valid informed consent because of impaired consciousness, severe cognitive impairment, or another reason, and no ethics-approved proxy consent procedure is available for this study.
• The participant declines research use of the medical record, endoscopic, pathological, imaging, or other required clinical data.
• Any other condition that, in the investigator's judgment, could seriously affect participant rights, data compliance, or the reliability of the study results.